Vitalik Buterin Backs AI Model That Deliberately Cut Its Hacking Skills
Ethereum co-founder Vitalik Buterin has endorsed Moonshot AI's decision to deliberately limit the hacking capabilities of its new Kimi K3.1 Pro model, arguing that restricting AI functions is preferable to restricting users.
The Beijing-based AI startup plans to release the model's open weights by November 4, allowing anyone to download and run it on their own hardware. Moonshot removed cyber-related data from its training corpus while keeping general coding tasks intact.
Performance Trade-Off
The safeguard comes with a measurable cost. Moonshot reports that the filtered model solved 81% of tasks on its internal exploit benchmark, while an unfiltered version solved 90.5%. DeepSeek's rival model V4.1-Flash already scored 85.7% on the same benchmark.
Race Against Time
Buterin warned that AI-powered attacks will eventually reach full force. He emphasized that developers need time to harden systems, check legacy code, and verify new software. Earlier this year, Binance Research found that AI exploits smart contracts roughly twice as well as it detects flaws.
Discriminating against use cases is MUCH less bad than discriminating against users. And the world does need all the time it can get to prepare, harden, check old…
Alternative Scenario
Buterin also warned about the alternative approach. If initial safeguards fail, he expects locked-down models, Know Your Customer (KYC) checks on everything, and intrusive surveillance. In that scenario, AI access could depend on credentials and nationality, potentially creating tiered user classes.
Related
